A digital predictive on-line energy optimization scheme for dc-dc converters

This work presents a novel energy conservation technique based on predicting the load current of a DC-DC converter that may feed a variety of loads, such as speakers and displays. The predicted load cnrrent is used to dynamically adjust the size of the output stage transistors and to switch into PFM mode to maximize the instantaneous converter efficiency. By using a segmented output stage, the trade-off between the gate drive losses and RMS conduction losses can be continuously optimized over the full load current range. The technique relies on the fact that the digital data stream which feeds modern electronic loads can be processed in real-time to predict the load cnrrent without relying on explicit cnrrent sensing or slow steady-state calibration techniques. The experimental prototype includes a digitally controlled 3.6V-to-1.8V DC-DC converter with an integrated segmented power stage IC operating at 4 MHz. A high-fidelity class-D audio amplifier acts as the DC-DC converter load. The results show a good agreement between the digitally predicted and actual DC-DC converter load current. The total energy consumed for three music pieces was reduced by up to 38% using the automatic mode/segment control technique. The fully digital efficiency optimization technique is well suited to future monolithic integration in advanced CMOS processes.